Show HN: I made a heatmap diff viewer for code reviews

Hot code map wins fans — until someone suggests scoring coworkers

TLDR: An open-source tool turns code reviews into an AI-colored heatmap, helping teams spot risky changes fast. Fans say it tames giant pull requests, while debate rages over UI tweaks and a spicy proposal to score how “easy to review” teammates’ work is — helpful feedback or gamified surveillance?

Developers are buzzing over a new AI-powered heatmap that turns code changes into a “hot spot” map for what needs a second look. It’s open source and dead simple to try: swap “github.com” with “0github.com” on any pull request (that’s the page where teammates approve code). Instead of hunting blindly, you get a color guide to the risky bits — think hard-coded secrets, funky encryption choices, or tangled logic. One fan joked this kills the lazy “LGTM” (internet-speak for “looks good to me”) reflex, saying the heatmap now “guides my eyes.”

But the real fireworks? A suggestion to add a checkbox that forces reviewers to answer the AI’s concerns — and start tracking how “easy to review” each teammate’s changes are. Cue the workplace scoreboard panic. Some cheered the accountability; others heard the faint hiss of corporate gamification. UI drama also flared: one user begged to ditch the slider and just let them click colored zones with labels. Another wanted it to learn each person’s review style, a nod to the flood of AI-generated code swamping teams. And the hipsters of hacker memory chimed in with a “seen it before,” dropping a link to salience. Love it or side-eye it, the crowd agrees: this thing makes giant code reviews a whole lot less painful.

Key Points

  • An open-source heatmap diff viewer highlights code changes by estimated need for human attention.
  • Users can try it by replacing “github.com” with “0github.com” in any GitHub pull request URL.
  • The system clones the repository into a virtual machine to process diffs.
  • Each diff is analyzed by an LLM (gpt-5-codex), which outputs a JSON data structure.
  • The JSON output is parsed into a colored heatmap overlay emphasizing areas worth a second look (e.g., secrets, crypto modes, complex logic).

Hottest takes

collecting stats on how "easy to review" PR's of team members are — jtwaleson
I could just click the different heatmap colors — timenotwasted
now the heatmap guides my eyes — austinwang115
Made with <3 by @siedrix and @shesho from CDMX. Powered by Forge&Hive.